摘要
为探明钢中不同形态及粒径的析出相粒子对钢的力学性能影响规律,需对其进行精确统计。为此,文章提出了一种基于形态特征和神经网络进行析出相自动分类统计的方法。该方法首先对目标图像进行预处理并提取目标粒子6个形态特征构成特征矢量以详细描述目标。然后利用BP神经网络建立粒子特征矢量与粒子形态的映射关系,继而实现对各种形态析出相粒子的自动分类统计。实验结果表明,该方法对诸如粒子团聚、粒子空洞及毛刺等缺陷目标具有很好的处理效果,可高效、便捷地进行析出相的自动分类,为钢中析出相的定量微观分析提供了可靠依据,而且具有很好的普适性。
In order to search that precipitated phase particle in alloyed steel has effect on mechanical property of alloyed steel, which have various shape and grain size, it is necessary to measure precipitated phase particle exactly. Automatic classification method of precipitates in steel based on morphological features and neural networks is proposed. After pre-processing of the image, six morphological features of precipitates was abstracted and used to describe the precipitates accurately. Then mapping the relationship between morphological features vector and shape of precipitate through BP neural network to achieve the purpose of automatic classification statistic of precipitates. The experiment results show that, automatic classification method is excellent to solve the problem such as precipitate's conglomeration, precipitates's holes and precipitates's burrs. Otherwise, the method can work efficiently and conveniently, it provides compellent evidence to the quantity analysis of precipitates in steel, and also can used in many other scientific fields.
出处
《塑性工程学报》
CAS
CSCD
北大核心
2009年第2期197-202,共6页
Journal of Plasticity Engineering
基金
国家自然科学基金资助项目(50775102)
江苏大学模具科技创新资助项目
关键词
析出相
形态特征
神经网络
自动分类
precipitates
morphological features
neural network
automatic classification